möglich sobald bei der ZB eingereicht worden ist.
VirNet: Deep attention model for viral reads identification.
Proceedings - 2018 13th International Conference on Computer Engineering and Systems, ICCES 2018, 623-626 (2019)
Metagenomics shows a promising understanding of function and diversity of the microbial communities due to the difficulty of studying microorganism with pure culture isolation. Moreover, the viral identification is considered one of the essential steps in studying microbial communities. Several studies show different methods to identify viruses in mixed metagenomic data using homology and statistical techniques. These techniques have many limitations due to viral genome diversity. In this work, we propose a deep attention model for viral identification of metagenomic data. For testing purpose, we generated fragments of viruses and bacteria from RefSeq genomes with different lengths to find the best hyperparameters for our model. Then, we simulated both microbiome and virome high throughput data from our test dataset with aim of validating our approach. We compared our tool to the state-of-the-art statistical tool for viral identification and found the performance of VirNet much better regarding accuracy on the same testing data.
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Artikel: Journalartikel
Schlagwörter Attention Model ; Classification ; Deep Neural Networks ; Metagenomics ; Virus
Konferenztitel Proceedings - 2018 13th International Conference on Computer Engineering and Systems, ICCES 2018
Quellenangaben Seiten: 623-626
Institut(e) Institute of Virology (VIRO)